Waste Plastic Identification With Thermal Barcodes: A Method That Also Sees Black Packaging

A US research team has developed a technique that identifies the polymer type of a plastic item using just a handful of infrared pulses – without contact, from a distance of some 20 centimetres. Crucially, it also works on black plastics, one of the biggest blind spots of today’s sorting plants. The study was published open access in Communications Engineering, part of the Nature Portfolio.

The problem: we do not know what we are sorting

The real bottleneck in plastics recycling is not the absence of recycling technology but the imprecision of identification. The authors state the global picture bluntly: only around 9 percent of the world’s plastic waste re-enters the loop, and recycling is in practice confined to PET (resin code 1) and HDPE (resin code 2). Higher resin code plastics are largely not recycled simply because no sorting technology is both accurate enough and cheap enough to handle them.

The consequence is quality loss. When polyethylene contaminates a polypropylene stream – or the reverse – the mechanical properties of the regranulate deteriorate, and the secondary material immediately loses its price advantage over virgin polymer. This is the cross-contamination problem that also directly threatens compliance with the EU’s PPWR recycled-content quotas: a legal obligation to use recyclate is of little use if the available material is not of consistent quality.

The authors also note that in US material recovery facilities (MRFs), plastic separation still relies heavily on manual picking. That produces frequent misidentification and exposes workers to the additives in the materials and to the hazards of contaminated waste.

How it works: microwave-oven logic at the molecular scale

The method is called the Transient Thermal Barcode (TTB), and its physics is refreshingly intuitive.

Think of a microwave oven. It heats food because water molecules absorb precisely the frequency the appliance emits, and the absorbed energy turns into heat. The same thing happens with plastics – except the energy is mid-infrared light rather than microwaves, and the targets are the chemical bonds of the polymer chain rather than water molecules.

Every chemical bond has a characteristic vibrational frequency. When a tunable infrared laser hits exactly that frequency, the plastic warms up at that spot – by a fraction of a degree, but measurably so. If the wavelength matches no bond, essentially nothing happens.

The key insight is that this warming can be observed from a distance with a thermal camera. The laser is focused to a spot roughly 1 millimetre across, a transient hotspot forms there, and the thermal camera captures it. Because plastics are poor heat conductors, the hotspot does not immediately dissipate, leaving time for imaging.

The experimental setup: a quantum cascade laser tunable between 5.2 and 12.8 micrometres, two gold mirrors, a zinc selenide focusing lens and, on the detection side, a FLIR thermal camera fitted with a polarizer that suppresses stray reflections from the plastic surface so that only the genuine, molecular-origin thermal signal passes through. Standoff distance was 20 centimetres.

When the team scanned across the full wavelength range, the resulting “thermal spectrum” reproduced essentially the same peaks as ATR-FTIR, the laboratory reference technique. In other words, a thermal camera can read the same chemical fingerprint remotely and without contact that a benchtop spectrometer reads by touching the sample.

Six wavelengths, six plastic types

Scanning the whole range, however, takes minutes – unusable next to a conveyor belt. This is where the paper’s real contribution lies.

The researchers asked what the minimum set of wavelengths is that still unambiguously separates the six main plastic types. The answer: six wavelengths suffice. Illuminate an item with those, and each polymer responds with its own characteristic heating pattern – a barcode made not of black bars but of hotspots of differing intensity.

Some examples: PET responds most strongly around 1463 cm⁻¹, the polyethylenes (HDPE, LDPE) show stronger absorption toward lower wavenumbers, and polystyrene produces a pronounced feature near 1376 cm⁻¹. What matters is not the absolute temperature rise but the ratio between the six values – and according to the measurements, that ratio is characteristic and repeatable for each polymer, even when samples come from different manufacturers and different products.

This ratio-based logic brings an important practical benefit: the method is self-referencing. Thicker samples produced broader, smoother peaks than thin films, but the peak positions did not shift. There is no need to calibrate a PET bottle separately from a PET tray.

What survived the lab: contamination, labels, blackness

Sorting plants do not handle clean laboratory samples, so the authors examined three realistic interferences.

Contamination. On samples coated with thin layers of coffee, tea, milk and oil, the characteristic polymer signal did not disappear; only its amplitude shifted slightly. Given the food residues that persist in collected packaging waste, this is a critical result.

Labels. Here lies the first serious limitation. The measurement always “sees” the topmost exposed material. If a label fully covers the measured spot, the system identifies the label, not the bottle. The researchers removed labels for their measurements. In a real installation this means scanning must cover multiple points, or the identification has to work in tandem with machine vision.

Black and dark plastics. This is the result that makes the study worth attention. Near-infrared (NIR) sorting machines – the backbone of European and Hungarian sorting plants – are notoriously blind to carbon-black pigmented plastics: the pigment absorbs the light and no usable signal returns. With the thermal barcode, however, the signal is the heat, and black pigment actually enhances infrared absorption.

In the tests, black variants showed different signal amplitudes, but peak positions and relative ratios remained comparable to non-black samples. The authors are candid that pigment-driven contrast differences require calibration and that a systematic treatment of this is left to future work; as a practical fix they propose using a conventional RGB camera to flag black items and applying separate thresholds to them.

Where the technology is not yet: speed

This part deserves a sober reading. The demonstrated apparatus illuminates the sample with the six wavelengths sequentially, not simultaneously, using a 100 millisecond dwell time per wavelength. The thermal camera runs at 30 frames per second because the uncooled microbolometer detector has a response time of 15–25 milliseconds. Excellent in a laboratory; not viable alongside a conveyor running at 3 metres per second.

The authors themselves set out the routes forward:

• Simultaneous illumination – a purpose-built source could emit all six wavelengths at once, with multiple regions of interest read in parallel across the full belt width.
• A cooled detector – an indium antimonide photon detector offers nanosecond-scale response and sub-20 millikelvin thermal sensitivity. The trade-off is that coolers have finite service life and need periodic maintenance, whereas microbolometers run for years untouched. A classic operations compromise.
• Simplified MEMS-based sensors – microcantilever devices are faster, small and mass-producible, so they could bring costs down. Focusing, signal amplification, mechanical stability and calibration standardisation remain to be solved.
• Mid-infrared LEDs in place of the expensive quantum cascade laser.

For industrial deployment the authors envisage scanning optics traversing the belt width, with identification results sent directly to the compressed-air jets and mechanical diverters already installed in MRFs. That is the best news in the whole paper: the approach does not demand a new sorting plant, only a new sensor on an existing line.

What this could mean for a Hungarian sorting plant

Viewed through a Hungarian lens, three points stand out.

First, black and dark packaging is not a theoretical problem here either. Black CPET trays and dark cosmetics and household bottles are effectively invisible to NIR sorters, so they typically end up in the residual fraction – energy recovery or landfill. Any technology that returns this fraction to material recycling directly improves the recycling performance of the Hungarian extended producer responsibility system.

Second, the PPWR recycled-content obligations are not about collected tonnage but about usable secondary material. Technology that reduces cross-contamination feeds straight into quota compliance – and on this point the study connects to recent work quantifying the effect of polyolefin cross-contamination on mechanical recycling.

Third, within the Hungarian concession model, sorting plant investments must clear a payback threshold. A sensor that retrofits into an existing line, rather than requiring the replacement of a whole process chain, has a far better chance of reaching real operation than a greenfield solution.

What to treat with caution

A few words for the critical reader, since the authors themselves claim no more than their results support.

This is laboratory proof of principle, not industrial validation. The paper reports no classification accuracy figures on a real mixed waste stream, no throughput data, and no cost comparison against deployed NIR systems – the last point the authors flag explicitly, stating that the method implies no automatic cost advantage over established visual or NIR/SWIR sorting.

The work is also confined to the six basic resin codes. Multilayer and composite packaging – among the most problematic components of collected lightweight packaging waste – was not among the samples, and the logic of the measurement (topmost layer wins) implies that this method alone cannot handle multilayer materials. Engineering plastics such as ABS and polyamide were likewise excluded.

Finally, access to the measurement data is only available on request from the authors, which limits independent verification.

Conclusion

The thermal barcode is not the kind of announcement that transforms sorting plants tomorrow. Its merit is that it takes a long-known physical phenomenon – localised heating from infrared absorption – and packages it into a form that works without contact, from a distance, using only a few wavelengths, and that is strongest precisely where today’s industrial standard is weakest: on black and contaminated plastics.

The next step is engineering rather than science: simultaneous illumination, faster detectors, realistic belt speeds, measurable sorting purity. If those come together, the technology can be introduced by swapping the sensor on an existing sorting line – and sorting accuracy is exactly the point at which the circular plastics economy has the most to gain.

Frequently Asked Questions About Waste Plastic Identification

How is waste plastic identified at sorting plants?

Waste plastic identification today relies mostly on near-infrared (NIR) sensors combined with manual picking. The NIR unit illuminates each item on the belt and infers the polymer type from the reflected signal. It is fast, but it frequently fails on dark and contaminated material, while manual sorting is slow and error-prone.

Why can’t black plastic be recycled?

The problem is detection, not the polymer itself. The carbon black that gives the plastic its colour absorbs near-infrared light, so the sorting machine receives no usable reflected signal – the item is effectively invisible to it. Black packaging therefore usually ends up in the residual fraction, headed for energy recovery or landfill.

How does thermal barcode plastic identification work?

The thermal barcode illuminates the plastic with infrared light. When the wavelength matches the vibrational frequency of a chemical bond, that spot heats up measurably, and a thermal camera captures the hotspot from a distance. The combined heating pattern across six well-chosen wavelengths identifies the polymer type unambiguously.

Why does the thermal barcode work on black plastic?

Because the measured signal is heat, not reflected light. Black pigment does not suppress infrared absorption – it enhances it, so the heating is actually stronger. In the study, black samples showed different signal amplitudes from coloured ones, but the pattern and the peak positions remained comparable.

What is a resin identification code?

The resin identification code (RIC) is the number from one to six inside the triangle on packaging, marking the base polymer: 1 – PET, 2 – HDPE, 3 – PVC, 4 – LDPE, 5 – PP, 6 – PS. The purpose of sorting is precisely to separate these fractions cleanly from one another.

What percentage of plastic is recycled globally?

According to the cited study, roughly 9 percent of the world’s plastic waste re-enters the loop. Recycling is in practice confined to resin codes 1 (PET) and 2 (HDPE) – for higher-numbered plastics there is simply no sorting technology that is both accurate enough and cheap enough.

Study details

Singh, K., Thundat, T. & Goyal, A.: Three-dimensional transient thermal barcode for waste plastic identification. Communications Engineering 5, 119 (2026). Published: 6 June 2026 | DOI: 10.1038/s44172-026-00703-7 Institution: University at Buffalo, Department of Chemical and Biological Engineering (New York, USA) Funding: New York State Center for Plastics Recycling Research and Innovation / New York State Department of Environmental Conservation Open access (CC BY-NC-ND 4.0): https://www.nature.com/articles/s44172-026-00703-7